Depends on the context, there is a lot of work in the scientific methods community trying to use NLP to augment traditionally fully human processes such as thematic analysis and systematic literature reviews and you can have protocols for validation there without 100% human review